**** BEGIN LOGGING AT Wed May 02 02:59:58 2012 May 02 04:58:32 can i use opnembedded to build a cross native compiler on my dm8168 arm board? May 02 06:10:17 <_tasslehoff_> Function failed: Fetcher failure for URL: 'file://var-run.conf'. Unable to fetch URL from any source <- this from base-files_3.0.14.bb, that hasn't changed since january and doesn't mention a var-run.conf. May 02 06:10:32 * _tasslehoff_ is confused May 02 06:12:48 maybe some .bbappend does May 02 06:20:35 good morning May 02 06:21:33 * _tasslehoff_ is unconfused, but a bit ashamed :) May 02 06:31:22 hm, meta-oe head does not work with yocto 1.2 :-( May 02 06:31:46 use meta-oe/denzil May 02 06:33:07 <_tasslehoff_> a case of two .bbappends setting FILESEXTRAPATHS := May 02 06:39:35 JaMa: ah, thanks May 02 06:40:38 btw anyone knows if there is a (non-classic) recipe for fastcgi ? does not seem to be in yocto 1.2 or meta-oe May 02 07:59:50 <_tasslehoff_> if more than one .bbappend supply a file, how do I control which one is chosen? May 02 08:00:22 <_tasslehoff_> does the order of paths in FILESEXTRAPATHS matter? May 02 08:00:49 yes it does May 02 08:33:32 hi.. i did "bitbake -f firefox" and then i want to rebuild my images, so i call "bitbake" but i get: ERROR: Nothing to do. Use 'bitbake world' to build everything, or run 'bitbake --help' for usage information. May 02 08:37:42 ah i need to run "bitbake myimage" .. May 02 08:38:12 tz.. holidays ;) May 02 08:44:13 morning all May 02 08:44:39 good morning May 02 08:45:50 anyone used fullscreen extension for firefox? May 02 10:12:06 hi all May 02 10:26:12 hi pb_ May 02 10:27:25 hi pb_, florian, all May 02 10:27:59 hi mckoan May 02 11:22:09 so, koen rejoins the tsc. that's a bit of a shame. May 02 11:23:32 why? May 02 11:42:21 hum well... May 02 11:42:40 JaMa: he is not good for PR May 02 11:43:10 PR is the boards job :-D May 02 11:43:41 well true, but the tsc acts in the public as well, controls resources etc. May 02 11:50:47 hm, some of us didn't even noticed, that he has quit the tsc :p May 02 11:58:45 florian: but he is very good for patches right? so from my POV it's better to have good devs then good PR May 02 12:04:51 well, one can send patches without being on the TSC. May 02 12:14:50 but who is better to discuss technical issues then good devs? as XorA said PR should be done by board not TSC May 02 12:24:27 just don't let him to do things "angstrom-centric" :) May 02 13:00:37 hello. I want to add some kernel config customizations to a custom layer May 02 13:00:52 I tried the .bbappend approach, but my defconfig is being ignored by bitbake May 02 13:01:17 I also noticed "KERNEL_FEATURES", and wonder what it really is. is there a special way for customizing the kernel config? May 02 13:17:19 * broonie finds himself re-adding stuff that was already present in oe-classic to try to boostrap mea-oe. May 02 13:17:22 meta-oe May 02 13:17:35 On the bright side, there's a little bit more documentation than there is for angstrom May 02 15:31:16 is there a clean way (except going manually into the source cache dir and pruning files there) how the downloaded files from SRC_URI can be removed? E.g. I have trouble with tcf-agent whose license text has been changed recently and the old file is lingering around at a lot of installations. May 02 15:32:32 something like -c cleanall? May 02 15:33:29 or you need to do it automatically with each "upgrade" build? May 02 15:34:51 JaMa: thx, cleanall seems to be enough May 02 15:35:40 that reminds me, we really need to get file:// files from SRC_URI added to the appropriate checksums. otherwise you're forced to manually cleansstate when changing them. worse yet, if using SSTATE_MIRRORS, there's no way to tell it not to re-fetch the shared state archive May 02 15:38:28 kergoth: are checksums of e.g. http:// files now part of sstate hash? May 02 15:38:47 no idea May 02 15:39:01 i doubt it, though, but that would be good too May 02 15:39:22 we've talked about this before, its just i happened to get bitten by it while hacking on external-csl-toolchain the other day (modified SUPPORTED) May 02 15:39:26 heh May 02 15:39:50 yup especially with PRSERV stuff (if it will end manuall PR bumps) May 02 15:40:18 good point May 02 15:40:28 hmm, would think it wouldn't be hard to do, actually.. May 02 15:41:12 add a ConfigParsed handler to checksum file:// files and store those in a variable or variables, add that to SRC_URI vardeps, and similarly suck in the various SRC_URI md5sum/shasum flags May 02 15:41:22 hm May 02 15:41:38 probably good to add it soon in release cycle as it will invalidate whole sstate cache :) May 02 15:41:56 * kergoth 'll try doing a quick prototype and see how it behaves May 02 16:36:14 * kenws is looking at the 64bithack.patch that gets applied to the GCC and wonders why --disable-multilib isn't used instead May 02 16:38:34 The patch seems strange. In case someone would enable multilib the locations for the m32 and m64 libs would be the same May 02 16:38:46 khem: RP: Any hints? : ) May 02 16:40:01 I've had a request from a customer of mine.. they would like -more- verbose bitbake information from the build.. basically when a task finishes, they'd like the contents of log file printed to the screen.. May 02 16:40:05 (even better if it was live) May 02 16:40:21 is there any way to do this, or would this be a custom UI? If it's a custom UI, any pointers to how to do this work? May 02 16:40:54 fray: what about the -v switch to bitbake? May 02 16:41:24 fray: I'm having a hard time imagining that being usable with multiple threads running, but I'm sure it would be possible... May 02 16:41:26 indeed or -DDD May 02 16:41:42 -DDD isn't it.. May 02 16:41:46 ok May 02 16:41:54 they don't want more spew from bitbake.. they want the logs of the build in context May 02 16:42:00 ah ok May 02 16:42:41 bluelightning I don't disagree.. the use-case specifically is "I want to do bitbake -c compile rpm" and I want the logs printed to the screen so I can immediate see what worked or didn't and inspect them.. they don't want to have to go digging through the tmp.../work/... directory structure to find the right files.. May 02 16:42:53 this also allows them to see an overall build captured in a single log file for grepping.. May 02 16:42:59 I'd certainly not advocate it be a default UI May 02 16:43:26 (I also suspect using this, you'd likely turn off parallel bitbake tasks.. but it could still be useful w/o it) May 02 16:44:01 OK... it kind of sounds like something that could be implemented as a command line option to the default ui May 02 16:44:46 that is what I was thinking.. but I don't have any experience with the UI to be able to do that.. May 02 16:45:05 I'm really not sure fi they prefer "live" listing, or when the task is complete (successfull or not) it just dump the results May 02 16:45:09 I'm not even sure if live is a possibility May 02 16:47:29 BTW what is the default UI? May 02 16:49:47 fray: knotty May 02 16:50:03 ok.. I think I see where the error logs are printed May 02 16:50:32 I see it checking bb.build.TaskFailed.. is there a TaskSucceeded? May 02 16:51:37 apparently there is, it's listed in the uihelper.py... May 02 16:53:21 Hi. Is anybody having problem with log4j1.2-native? I'm getting this error: http://paste.debian.net/167021/ May 02 16:54:13 fray: look at knotty.py, its straight forward to add something in there which would cat the log at task completetion May 02 16:54:35 fray: since its the UI that writes to the screen, the multiple process problem shouldn't be an issue in that case May 02 16:55:17 kenws: We build a toolchain per multilib and this was the simplest way to avoid some nasty logic figuring things out May 02 16:55:34 RP: I think the reason for the 64bithack.patch is the libstdc++. With --disable-multiarch the libstdc++ would still be placed in lib64 which is not what you want. May 02 16:55:37 kenws: we are working on something better but its not been ready for master yet May 02 16:56:02 kenws: there are a variety of bad things that code can do May 02 16:57:01 RP: I think deleting the MULTILIB_OSDIRNAMES MULTILIB_DIRNAMES variables from the i386/t-linux64 should also achive this May 02 16:57:52 This causes the GCC to support just one flavour (m32 or m64) even if you enable multilib May 02 16:57:55 kenws: ultimately we're going to write out the multilib configuration into there May 02 16:58:57 RP: Hm, I'm afraid I don't get it. Are you saying that you want oe to support multilib in the future? May 02 17:00:25 kenws: we support multilib now with a toolchain per multilib. This doesn't let us build a toolchain for use on target that will support multilib though May 02 17:01:07 RP: alright, so far I'm with you May 02 17:01:59 kenws: if we fix the gcc configuration to match the full set of multilibs, it will work on target which is a good idea IMO May 02 17:02:07 kenws: sorry, I need to go :( May 02 17:02:40 RP: ok, thanks so far. Maybe you could check the irc log later again May 02 17:04:50 bluelightning how does hob watch the task log directly in the UI? May 02 17:05:36 fray: AFAIK it works in the same way as knotty, by listening for events May 02 17:06:03 yes.. but what actually watches the logs.. May 02 17:06:13 since they appear to be displayed live, and not on the succeeded or failed event May 02 17:06:32 my assumption is that something captures the event.logfile on task start.. adds it to the tree.. May 02 17:06:39 but I'm curious how it's being run... May 02 17:07:11 I wasn't aware that it actually looked at the log for a task unless it failed May 02 17:07:35 in the UI on hob, I swear you get a tree view of all of the tasks and you can expand and watch them... live even May 02 17:08:06 yeah, but that's just a list of what tasks have run and are running; that's solely based on the events May 02 17:08:22 yes.. it's the printing of hte logs part I'm wondering how it's implemented May 02 17:08:46 if it is printing the actual logs of individual tasks that's functionality I haven't seen May 02 17:08:50 since I can do the same event processing in the knotty ui (or any other ui).. but I don't know how to watch a file "live" based on the events.. I do see how to do it after the task has finished though May 02 17:10:27 I can't run hob unfortunately, or I would run it and verify what I remember May 02 17:10:45 (hob complains it can't run remote due to TCP/IP netowrking for ORBit) May 02 17:15:13 hmm, that's a bit sad May 02 17:18:32 hey if I manually run dbus it starts to work! May 02 17:19:35 fray: actually you're right, it does include the task output in the tree May 02 17:19:48 fray: maybe the code is in runningbuild.py? May 02 17:20:29 I note that most of the ui/hob.py code seems to just be very basic... where is the majority of the implementation? May 02 17:21:43 Hmm the new hob doesn't see to let you see the log live, only once the task is done May 02 17:22:05 is hob1 still available? May 02 17:22:25 fray: in the git history only May 02 17:22:28 AFAIK May 02 17:22:40 ok.. I can back off to a prior release and double check this May 02 17:24:15 Hmm.. scrolling through the logs is -really- slow.. (I'm on a remote x-term.. so I don't know if that is an artifact of being remote or not) May 02 17:24:32 (remote over an 802.11g link) May 02 17:24:53 says 54Mbit/s rate.. ;) May 02 17:25:31 ah, well latency could be the issue there May 02 17:25:50 I suspect gtk isn't optimizaing the scrolling.. but actually trying to display -everything- May 02 17:26:30 X11-over-ssh performance on a local LAN with machines right next to eachother is pretty poor if you ask me, let alone wireless May 02 17:26:48 ya.. thats what I'm stuck with.. the amchines are in the server lab next door May 02 17:28:02 ok I gotta run, may be back online later May 02 17:56:24 RP: khem: I think what the GCC folks are saying is that even if you disable multilib the libstdc++ will be placed in lib64 during the install step. Because that's where 64bit libs go for linux targets according to the FHS. If you don't like this but want to use linux as a target you have to patch the toolchain which is what 64bithack.patch attempts to do. One solution would be to have a distinctive OE target that doesn't support multilib and has 64bit libs under lib May 02 17:56:24 . Another hack would be to just delete the MULTILIB_* variables from the gcc/config/i386/t-linux64. The advantage over the current 64bithack.patch would be that no one could ever build a broken multilib toolchain. Even easier, we could just this: http://paste.ubuntu.com/962860. May 02 17:56:43 florian: which channel? May 02 17:57:19 I'm currently testing this patch against meta-linaro and it's looking good May 02 18:00:07 mickeyl: #oe-board is the one we used the last time iirc May 02 18:01:56 k May 02 21:14:18 Hey all. I'm trying to include a custom kernel module in a custom image (based on omap3-console-image, and I'm getting an error resolving conffiles. Apparently an existing conffile (that is different from the new one) already exists. What does this mean, in general terms? (I've been googling, but I'm still confused.) Thanks in advance. May 02 22:08:49 JeremyClarkson: which conffile ? May 02 22:09:18 khem: device_table May 02 22:15:29 hmmm are you using latest oe May 02 22:15:37 this issue was dealt with in angstrom May 02 22:15:42 iirc anyway May 02 22:16:04 I beleive i May 02 22:16:14 I believe I'm using the 2011.3 branch May 02 22:16:25 oh hmm May 02 22:19:59 I suspect my do_install task is bad. I tried removing it (to use the default module task, but it wouldn't build) May 02 22:27:35 check if there are other samples of external modules in tree May 02 22:27:38 and follow the steps May 02 22:28:16 ok May 03 00:10:45 kenws: multilibs are not consistent in gcc May 03 00:11:05 e.g. some 64 arches use lib as default some use lib64 May 03 00:11:29 OE does not suffer from that since we have our own notion of multilib and its consistent May 03 00:12:39 kenws: on another note you can now take my patches for meta-linaro now that gcc 4.7 is merged and I see that you have tagged denzil **** ENDING LOGGING AT Thu May 03 03:00:01 2012